land-use and land-cover change carbon emissions between 1901 … · 2017. 11. 14. · 2011; saatchi...
Post on 31-Aug-2020
0 Views
Preview:
TRANSCRIPT
Biogeosciences, 14, 5053–5067, 2017https://doi.org/10.5194/bg-14-5053-2017© Author(s) 2017. This work is distributed underthe Creative Commons Attribution 3.0 License.
Land-use and land-cover change carbon emissions between 1901and 2012 constrained by biomass observationsWei Li1, Philippe Ciais1, Shushi Peng1,2, Chao Yue1, Yilong Wang1, Martin Thurner3, Sassan S. Saatchi4,Almut Arneth5, Valerio Avitabile6, Nuno Carvalhais7,8, Anna B. Harper9, Etsushi Kato10, Charles Koven11, YiY. Liu12, Julia E.M.S. Nabel13, Yude Pan14, Julia Pongratz13, Benjamin Poulter15, Thomas A. M. Pugh5,16,Maurizio Santoro17, Stephen Sitch18, Benjamin D. Stocker19,20, Nicolas Viovy1, Andy Wiltshire21,Rasoul Yousefpour13,a, and Sönke Zaehle7
1Laboratoire des Sciences du Climat et de l’Environnement, LSCE/IPSL, CEA-CNRS-UVSQ,Université Paris-Saclay, 91191 Gif-sur-Yvette, France2Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences,Peking University, Beijing 100871, China3Department of Environmental Science and Analytical Chemistry (ACES) and the Bolin Centre for ClimateResearch, Stockholm University, 106 91 Stockholm, Sweden4Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109, USA5Karlsruhe Institute of Technology, Institute of Meteorology and Climate Research – Atmospheric EnvironmentalResearch (IMK-IFU), Garmisch-Partenkirchen, Germany6Centre for Geo-Information and Remote Sensing, Wageningen University & Research, Droevendaalsesteeg 3,6708PB Wageningen, the Netherlands7Department for Biogeochemical Integration, Max-Planck-Institute for Biogeochemistry, Jena, Germany8CENSE, Departamento de Ciências e Engenharia do Ambiente, Faculdade de Ciências e Tecnologia,Universidade NOVA de Lisboa, Caparica, Portugal9College of Engineering, Mathematics, and Physical Sciences, University of Exeter, Exeter, UK10Institute of Applied Energy, Minato, Tokyo 105-0003, Japan11Climate and Ecosystem Sciences Department, Lawrence Berkeley Lab, Berkeley, CA, USA12ARC Centre of Excellence for Climate Systems Science & Climate Change Research Centre,University of New South Wales, Sydney, New South Wales 2052, Australia13Max Planck Institute for Meteorology, Hamburg, Germany14USDA Forest Service, Durham, New Hampshire, USA15Department of Ecology, Montana State University, Bozeman, MT 59717, USA16School of Geography, Earth & Environmental Science and Birmingham Institute of Forest Research,University of Birmingham, Birmingham, B15 2TT, UK17GAMMA Remote Sensing, 3073 Gümligen, Switzerland18College of Life and Environmental Sciences, University of Exeter, Exeter, UK19Climate and Environmental Physics, and Oeschger Centre for Climate Change Research,University of Bern, Bern, Switzerland20Imperial College London, Life Science Department, Silwood Park, Ascot, Berkshire SL5 7PY, UK21Met Office Hadley Centre, Exeter, Devon, EX1 3PB, UKacurrent address: Chair of Forestry Economics and Forest Planning, University of Freiburg, 79106 Freiburg, Germany
Correspondence to: Wei Li (wei.li@lsce.ipsl.fr)
Received: 12 May 2017 – Discussion started: 2 June 2017Revised: 12 October 2017 – Accepted: 12 October 2017 – Published: 14 November 2017
Published by Copernicus Publications on behalf of the European Geosciences Union.
5054 W. Li et al.: Land-use and land-cover change carbon emissions
Abstract. The use of dynamic global vegetation models(DGVMs) to estimate CO2 emissions from land-use andland-cover change (LULCC) offers a new window to ac-count for spatial and temporal details of emissions and forecosystem processes affected by LULCC. One drawback ofLULCC emissions from DGVMs, however, is lack of obser-vation constraint. Here, we propose a new method of usingsatellite- and inventory-based biomass observations to con-strain historical cumulative LULCC emissions (Ec
LUC) froman ensemble of nine DGVMs based on emerging relation-ships between simulated vegetation biomass and Ec
LUC. Thismethod is applicable on the global and regional scale. Theoriginal DGVM estimates of Ec
LUC range from 94 to 273 PgCduring 1901–2012. After constraining by current biomassobservations, we derive a best estimate of 155± 50 PgC(1σ Gaussian error). The constrained LULCC emissions arehigher than prior DGVM values in tropical regions but sig-nificantly lower in North America. Our emergent constraintapproach independently verifies the median model estimateby biomass observations, giving support to the use of this es-timate in carbon budget assessments. The uncertainty in theconstrained Ec
LUC is still relatively large because of the un-certainty in the biomass observations, and thus reduced un-certainty in addition to increased accuracy in biomass obser-vations in the future will help improve the constraint. Thisconstraint method can also be applied to evaluate the impactof land-based mitigation activities.
1 Introduction
Carbon emissions from land-use and land-cover change(LULCC) are part of the human perturbation to the globalcarbon cycle (Houghton et al., 2012; Le Quéré et al.,2015) and started before the industrial era when fossil fuelCO2 emissions appeared. Since 1850, estimated cumulativeLULCC emissions, Ec
LUC, have represented one-third of to-tal cumulative anthropogenic CO2 emissions (Boden et al.,2013; Houghton et al., 2012; Le Quéré et al., 2015). AnnualLULCC emissions have been higher than those from fossilfuel burning until the 1930s (Boden et al., 2013; Houghton etal., 2012; Le Quéré et al., 2015) and today represent a smallerbut persistent perturbation in the global carbon cycle. Unlikefossil fuel emissions, relative uncertainties in LULCC emis-sions are high due to the difficulty of assessing this flux frommeasurements. Some progress has been made to better quan-tify gross tropical deforestation emissions by combining spa-tial biomass data with satellite-derived maps delineating for-est cover loss (Harris et al., 2012). However, such spatiallyresolved data are not available beyond the last decade andprovide only gross deforestation emissions, i.e., do not trackthe regrowth of secondary ecosystems or legacy soil carbonlosses that can persist long after deforestation.
Bookkeeping models (Hansis et al., 2015; Houghton,1999) based on historical LULCC area data and tabulatedfunctions of carbon losses and gains are one approach to es-timating Ec
LUC, but they do not include the effects of environ-mental changes on carbon stocks before and after LULCChappens (Gasser and Ciais, 2013; Pongratz et al., 2014).The bookkeeping model of Houghton (1999) used for theannual update of the global carbon budget (Le Quéré etal., 2015) is based on regionally aggregated data and doesnot consider spatial differences in LULCC fluxes within aregion. Alternatively, the estimated LULCC fluxes by dy-namic global vegetation models (DGVMs) account for spa-tial and temporal variations in carbon stock densities andland-cover change, as well as for delayed (“legacy”) carbonfluxes. In DGVMs, LULCC fluxes are related to environmen-tal conditions through simulated carbon cycle processes, i.e.,net primary production (NPP) and respiration, resulting inchanges in biomass and soil carbon stocks simulated withvariable atmospheric CO2 concentration and climate. Yet,LULCC emissions from DGVMs differ greatly, even whenthese models are prescribed with the same inputs of land-cover change data (such as time-variable areas of pasture andcrops; Pitman et al., 2009). Several factors are responsible fordifferences in Ec
LUC among DGVMs, including (1) differentrepresentations of processes that determine the carbon den-sities of vegetation and soils subject to land-use change; (2)using dynamic vegetation or prescribing a fixed vegetationdistribution; and 3) the use of different rules assigning hownatural vegetation types change to agricultural areas (Peng etal., 2017; Pitman et al., 2009; Reick et al., 2013).
Carbon initially stored in forest biomass contributes thepredominant portion of the LULCC emissions after defor-estation (Hansis et al., 2015). Thus, an accurate represen-tation of the biomass carbon density exposed to LULCC iscrucial to reduce uncertainties in DGVM-based Ec
LUC es-timates. Global biomass datasets based on inventories andsatellites recently became available. These datasets (Table 1)provide the spatially distributed biomass carbon density onregional or global scales (Avitabile et al., 2016; Baccini etal., 2012; Carvalhais et al., 2014; Liu et al., 2015; Pan et al.,2011; Saatchi et al., 2011; Santoro et al., 2015; Thurner et al.,2014), but differ in terms of their coverage of aboveground orbelowground biomass and whether they provide only forestbiomass or biomass for all vegetation types.
In this study, we propose a new method to combine re-cent satellite- and inventory-based biomass datasets to con-strain Ec
LUC simulated by DGVMs (Fig. 1). We analyzed theoutputs from nine DGVMs (Table 2) of the Trends in NetLand–Atmosphere Exchange (TRENDY-v2) project (Sitchet al., 2015; http://dgvm.ceh.ac.uk/node/9) and developedglobal and regional regressions between initial biomass in1901 and present-day biomass (average of 2000–2012) andbetween Ec
LUC during 1901–2012 and initial biomass across
Biogeosciences, 14, 5053–5067, 2017 www.biogeosciences.net/14/5053/2017/
W. Li et al.: Land-use and land-cover change carbon emissions 5055
Forest area change in PFT maps
Regression between cumulative LULCC emissions and initialbiomass (in 1901) in models
Deforestation grid cells since 1901 in PFT maps
Cumulative LULCC emissions constrained by biomass observations
Present global biomass map based on observations
Regression between initial biomass (in 1901) and present biomass in models
Observation-based biomass in 1901 from Method A, B, C
Present biomass
Initi
al
biom
ass
Initial biomassLU
LCC
em
issi
ons
Present biomass in deforestation grid cells using Method A, B, C
Figure 1. The framework of this study.
the DGVMs. The former set of regressions is used to extrap-olate present-day observation-based biomass (Table 1) to ini-tial biomass in the year 1901. The latter set of regressions isapplied to provide an emerging constraint on Ec
LUC as a func-tion of initial biomass (Fig. 1). Using the Gaussian uncertain-ties associated with the observation-based biomass datasetsand the uncertainties in the two regressions, the Gaussian er-rors in Ec
LUC can be derived after applying the biomass con-straint.
2 Materials and methods
2.1 LULCC emissions and biomass from the DGVMs
The DGVMs in TRENDY-v2 was used to conduct two sim-ulations (labeled S2 and S3) between 1860 (except JSBACHfrom 1850, Table 2) and 2012, with outputs quantifyingLULCC emissions over the period 1901–2012 (Sitch et al.,2015). Both simulations are performed with changing cli-mate and CO2 concentration, but one (called S3) has variableLULCC maps based on Land-Use Harmonization (LUH)dataset (Hurtt et al., 2011; with an extension until 2012),and the other (called S2) has a time-invariant land-cover maprepresenting the state in 1860. The difference in net biomeproduction (NBP, the net carbon exchange between the bio-sphere and the atmosphere) between these two simulations(S3 and S2) defines modeled LULCC emissions. This cal-culation of LULCC emissions by DGVMs includes the “lostsink capacity” (called “altered sink capacity” in Gasser andCiais, 2013, and “the loss of additional sink capacity” in Pon-gratz et al., 2014) because simulated NBP in the S2 simu-lation without LULCC is a net sink over areas affected byLULCC in S3. For example, forests have larger carbon stor-
age and a slower turnover time than croplands and are thusexpected to be carbon sinks when the atmospheric CO2 levelincreases. After deforestation to croplands, this sink capac-ity due to CO2 fertilization is lost. Modeled LULCC emis-sions include the legacy emissions from soil carbon lossesand emissions from wood and other products produced byLULCC, as far as the latter are included in the TRENDY-v2 models (Table 2). The DGVMs used in this study areCLM4.5 (Oleson et al., 2013), JSBACH (Reick et al., 2013),JULES3.2 (Best et al., 2011; Clark et al., 2011), LPJ (Sitchet al., 2003), LPJ-GUESS (Smith et al., 2001), LPX-Bern(Stocker et al., 2014), ORCHIDEE (Krinner et al., 2005),VISIT (Ito and Inatomi, 2012; Kato et al., 2013) and OCN(Zaehle and Friend, 2010). Each DGVM is described brieflyin Table 2.
LULCC can either reduce or increase the biomass amountover time depending on the LULCC types. For example,forest clearing turns forest biomass into atmospheric CO2eventually, while secondary forest regrowth can increasebiomass. The overall effect of LULCC on biomass duringthe historical period is a net loss of carbon (Houghton, 1999)due to converting natural vegetation into cultivated landsby humans (Klein Goldewijk et al., 2011). Identifying theLULCC-affected grid cells in each model is thus critical be-cause only biomass in these grid cells should be used to con-strain LULCC emissions. Grid cells affected by LULCC dif-fer among models. Although all models share the same pas-ture and cropland areas from the LUH dataset (Hurtt et al.,2011), the models have different numbers of PFT, use dif-ferent PFT definitions and have different allocation rules fortranslating the shared agricultural data into the new vegeta-tion cover (Peng et al., 2017; Pitman et al., 2009; Reick et al.,2013). As a result, there is no unified map to determine the
www.biogeosciences.net/14/5053/2017/ Biogeosciences, 14, 5053–5067, 2017
5056 W. Li et al.: Land-use and land-cover change carbon emissions
Table1.T
hedifferentbiom
assdatasets
basedon
observations.The
biomass
information
fromthe
TR
EN
DY
-v2projectis
alsolisted
forcomparison.
Dataset
Coverage
Resolution
Biom
etype
Aboveground/below
groundN
ote
Thurneretal.(2014)
30◦
N–80◦
N0.01◦
forestaboveground
+below
groundG
rowing
stockvolum
efrom
Santoroet
al.(2015)Saatchietal.(2011)
30◦
N–40◦
S1
kmforest
abovegroundC
arvalhaisetal.(2014)
global(withoutSouth
Australia)
0.5◦
forest+
herbaceousaboveground
+below
groundM
ergedm
apof
Thurner
etal.
(2014)and
Saatchietal.(2011)B
accinietal.(2012)23◦
N–23◦
S500
mforest
abovegroundL
iuetal.(2015)
global0.25
degreeall
abovegroundC
alibrationbased
onSaatchietal.(2011)
Avitabile
etal.(2016)30◦
N–40◦
S1
kmforest
abovegroundFusion
ofSaatchietal.(2011)and
Baccinietal.(2012)
Santoroetal.(2015)
30◦
N+
0.01◦
forestaboveground
Sharinggrow
ingstock
volume
with
Thurneretal.(2014)
Panetal.(2011)
globalregional
forestaboveground
+below
groundB
asedon
FAO
dataT
RE
ND
Y-v2
global1◦
allaboveground
+below
ground
China region
0
5
10
North America
0
5
10
South and Central America
Western Europe
Tropical Africa
1910 19600
5
10
The former Soviet Union
South and Southeast Asia
1910 1960
Pacific developed region
1910 1960 2010
North Africa and Middle East
CLM4.5JSBACHJULES3.2LPJLPJ-GUESSLPX-BernORCHIDEEVISITOCN
For
est a
rea
(mill
ion
km)
2
Year
Figure 2. Temporal change in forest area from TRENDY-v2 modelsin each of the nine regions. Differences between models arise fromtheir specific vegetation maps and rules through which natural PFTsare chosen to give land to agriculture.
LULCC-affected grid cells in all models. For the same rea-sons, the forest areas and the LULCC types are also differentamong models.
In this study, we adopted the “deforestation grid cells” intheir corresponding PFT maps as a criterion to locate theLULCC-affected grid cells from DGVM outputs. Thus weused the PFT maps from each model to first calculate thetemporal change in forest area (total area of all forest PFTs)during 1901–2012 and then selected the grid cells that ex-perienced deforestation by comparing the forest area mapsbetween 1901 and 2012 (net deforestation). This procedureproduces a good approximation given the continuously de-creasing trend of forest area in LULCC hotspot regions likeSouth and Central America (Fig. 2). We also tested an alter-native method to determine the LULCC-affected grid cellsin TRENDY model outputs; i.e., PFT maps were comparedyear by year during 1901–2012, and grid cells with deforesta-tion were selected (gross deforestation). This method tends togive a greater number of LULCC-affected grid cells, reduc-ing the goodness of fit in the regression between the biomassin 1901 and Ec
LUC during 1901–2012 (Figs. S1 and S2 in theSupplement). Therefore, the method of gross deforestation isnot used for further analyses.
We verified that deforestation grid cells are responsible formost of the total net LULCC flux. In fact, the average ofthe different model simulations of LULCC emissions fromdeforestation grid cells between 1901 and 2012 is approxi-mately 90 % of the total LULCC emissions from all grid cells(Fig. S1). The LULCC emissions in this study are thus takento equal the sum of LULCC emissions from the selected de-forestation grid cells using our criterion. It should be notedthat although only deforestation is used as a single criterionto define grid cells affected by LULCC in DGVMs, modeledLULCC emissions also include other types of land-use tran-
Biogeosciences, 14, 5053–5067, 2017 www.biogeosciences.net/14/5053/2017/
W. Li et al.: Land-use and land-cover change carbon emissions 5057
Tabl
e2.
Des
crip
tion
ofT
RE
ND
Ym
odel
setu
psus
edin
this
stud
y.
Mod
elPF
Tnu
mbe
rA
lloca
tion
rule
sof
chan
ges
inag
ricu
lture
area
Spat
ial
reso
lutio
ndy
nam
icve
geta
tion
activ
ated
Woo
dha
rves
tSh
iftin
gag
ricu
lture
Exp
licit
gros
sL
UL
CC
tran
sitio
ns
Star
tof
tran
sien
tsi
mul
atio
n
Ref
eren
ce
CL
M4.
517
past
ure;
crop
mod
elex
ists
but
notu
sed
inth
ese
sim
ulat
ions
0.9◦×
1.25◦
noye
sno
no18
60O
leso
net
al.(
2013
)
JSB
AC
H12
prop
ortio
nala
lloca
tion
ofcr
opla
nd;
pref
eren
tiala
lloca
tion
ofpa
stur
eon
natu
ralg
rass
land
T63
ano
yes
yes
yes
1850
Rei
cket
al.(
2013
)
JUL
ES3
.25
crop
and
past
ure
adde
dto
geth
erto
crea
tea
sing
leag
ricu
ltura
lm
ask
inw
hich
tree
san
dsh
rubs
are
excl
uded
from
grow
ing;
ther
eis
noas
sum
ptio
nfo
rw
hich
PFT
sth
eag
ricu
lture
repl
aces
N96
bye
sye
sno
no18
60C
lark
etal
.(20
11),
Bes
teta
l.(2
011)
LPJ
9cr
opan
dpa
stur
ew
ere
adde
dto
geth
erto
crea
tea
sing
lem
anag
edla
ndfr
actio
n0.
5◦×
0.5◦
yes
nono
no18
60Si
tch
etal
.(20
03)
LPJ
-GU
ESS
11pr
opor
tiona
lallo
catio
nof
crop
land
and
past
ure
0.5◦×
0.5◦
yes
nono
no18
60Sm
ithet
al.(
2001
)
LPX
-Ber
n9
prop
ortio
nala
lloca
tion
ofcr
opla
ndan
dpa
stur
e1.
0◦×
1.0◦
yes
nono
no18
60St
ocke
reta
l.(2
014)
OR
CH
IDE
E13
prop
ortio
nala
lloca
tion
ofcr
opla
ndan
dpa
stur
e2◦×
2◦no
nono
no18
60K
rinn
eret
al.(
2005
)
VIS
IT16
nosp
ecifi
cru
leap
plie
dbe
caus
eon
lyon
ena
tura
lPFT
exis
tsfo
rpri
mar
yan
dse
cond
ary
land
ina
grid
cell
0.5◦×
0.5◦
noye
sye
sye
s18
60K
ato
etal
.(20
13),
Ito
and
Inat
omi(
2012
)
OC
N12
prop
ortio
nala
lloca
tion
ofcr
opla
ndan
dpa
stur
e2.
5◦×
3.75◦
nono
nono
1860
Zae
hle
and
Frie
nd(2
010)
aT
63gr
idha
san
appr
oxim
ate
reso
lutio
nof
1.9◦×
1.9◦
.bN
96re
solu
tion
iseq
uiva
lent
to1.
25◦
latit
ude×
1.87
5◦lo
ngitu
de.
www.biogeosciences.net/14/5053/2017/ Biogeosciences, 14, 5053–5067, 2017
5058 W. Li et al.: Land-use and land-cover change carbon emissions
sitions involving pairs of non-forest PFTs in the selected gridcells.
In each model, only biomass in deforestation grid cells isconsidered. Biomass in the year 1901 is thereby defined asinitial biomass, and biomass averaged during 2000–2012 isdefined as present biomass. An ordinary least squares lin-ear regression is performed with the outputs of all modelsbetween initial biomass and Ec
LUC from 1901 to 2012 andbetween the initial and the present biomass on both globaland regional scales. Our division of nine regions in the world(Fig. 2) for estimating LULCC fluxes is the same as inHoughton et al. (1999).
2.2 Observation-based biomass datasets
Several biomass datasets (Avitabile et al., 2016; Baccini etal., 2012; Carvalhais et al., 2014; Liu et al., 2015; Pan et al.,2011; Saatchi et al., 2011; Santoro et al., 2015; Thurner etal., 2014) based on inventories and remote sensing can po-tentially be used to constrain Ec
LUC through the set of re-gressions from DGVMs. However, these biomass datasetscover different parts of biomass (aboveground, belowgroundor total) and different regions (tropics, Northern Hemisphereor the globe) at different spatial resolutions (Table 1). Wechoose the global grid-based biomass dataset from Carval-hais et al. (2014) to derive an observational constraint that re-sults in a best estimate of Ec
LUC. This map merges the North-ern Hemisphere biomass dataset from Thurner et al. (2014)and the tropical biomass dataset from Saatchi et al. (2011).An advantage of this map is its consistency in biomassterms with the outputs of TRENDY models because it docu-ments aboveground+ belowground and forest+ herbaceousbiomass (Tables 1, 2). Three other biomass maps are usedas alternative datasets for sensitivity tests: (1) the globalbiomass map from the GEOCARBON project, a mergedproduct of the biomass datasets in the Northern Hemisphere(Santoro et al., 2015) and tropics (Avitabile et al., 2016);(2) regional biomass estimates from Pan et al. (2011) basedon forest inventory data; and (3) the biomass map from Liu etal. (2015) derived from satellite vegetation optical depth. TheGEOCARBON (Avitabile et al., 2016; Santoro et al., 2015)and Liu et al. (2015) datasets that only provide abovegroundbiomass were extended to total forest biomass using the con-version factors for the nine regions (Liu et al., 2015). Theglobal biomass maps from GEOCARBON (Avitabile et al.,2016; Santoro et al., 2015) and Pan et al. (2011) are only forforest (Table 1), and we do not add the herbaceous biomassto these two datasets because the global herbaceous biomassonly accounts for about 3 % of the global total biomass (Car-valhais et al., 2014). Note that the uncertainties in the cor-responding constrained results using these three alternativedatasets do not include (1) the uncertainties in convertingaboveground biomass to the total of aboveground and below-ground biomass for the datasets from Liu et al. (2015) andGEOCARBON (Avitabile et al., 2016; Santoro et al., 2015)
or (2) the uncertainties in ignoring non-woody biomass in thedatasets from GEOCARBON (Avitabile et al., 2016; Santoroet al., 2015) and Pan et al. (2011). The biomass maps of Car-valhais et al. (2014), GEOCARBON (Avitabile et al., 2016;Santoro et al., 2015) and Liu et al. (2015) with different spa-tial resolutions were aggregated to a 1◦× 1◦ resolution be-fore selecting the deforestation grid cells.
2.3 Methods to identify grid cells subject to pastdeforestation in biomass datasets
It is not practical to use PFT maps from DGVMs to de-fine deforestation grid cells in the observation-based biomassdatasets because PFT maps and forest area change since1901 differ across DGVMs. Instead, we diagnosed defor-estation grid cells in the biomass maps using three harmo-nized methods (Method A, Method B and Method C). Allthe methods are based on the reconstructed historical agri-cultural area from the History Database of the Global En-vironment (HYDE v3.1; Klein Goldewijk et al., 2011) butwith different hypotheses regarding how agricultural expan-sion has affected forests. These harmonized methods are rep-resentative of the different rules for assigning LULCC datato natural vegetation types in DGVMs. Method-A assumesthat the increase in cropland area in a grid cell between 1901and 2012 is taken from forest; Method B assumes that theincrease in cropland and pasture is taken proportionally fromall natural vegetation types; and Method C (like the “BM3”scenario in Peng et al., 2017) assumes that the increase incropland and pasture is first taken from forest and then fromnatural grassland if no more forest area is available and thatthe regional forest area change is set to match the histori-cal forest reconstruction from Houghton (2003). Because thebiomass distribution in Pan et al. (2011) is given as regionalmean values and not resolved on a grid cell basis, it is im-possible to select deforestation grid cells directly from thisdataset using the above methods. Therefore, for each region,we calculated the ratios of biomass in deforestation grid cellsaccording to Method A, Method B and Method C to the totalbiomass in all grid cells in each of the other three biomassdatasets (Carvalhais et al., 2014; GEOCARBON, Avitabileet al., 2016; Santoro et al., 2015; Liu et al., 2015). For eachmethod (Method A, B and C), the three ratios correspond-ing to the three biomass datasets were further averaged ineach region. The total biomass amount from Pan et al. (2011)in each region was multiplied by the average ratio to derivethe biomass equivalent to using Method A, Method B andMethod C for the dataset from Pan et al. (2011).
These three methods applied to the above-listed biomassdatasets are also applied as sensitivity tests to select the de-forestation grid cells since 1901 in the TRENDY model out-puts. Identically, regressions are performed using the initialbiomass amount and Ec
LUC from these selected grid cells.Due to the inconsistencies among the three methods and thehistorical PFT maps of each DGVM, the biomass amount in
Biogeosciences, 14, 5053–5067, 2017 www.biogeosciences.net/14/5053/2017/
W. Li et al.: Land-use and land-cover change carbon emissions 5059
1901 in the selected grid cells using these three methods ishigher than using PFT maps, but the Ec
LUC are lower, reflect-ing a lower representativeness of the deforestation grid cellsusing these three methods for DGVM outputs (Fig. S1). As aconsequence, a weaker goodness of regression fit was foundbetween Ec
LUC and initial biomass (Fig. S2).
2.4 Uncertainties in constrained LULCC emissions
The biomass from Method A, Method B and Method C ob-tained from each dataset is extrapolated into biomass forthe year 1901 using the regression between initial biomassand present biomass modeled by the DGVMs. This biomassin 1901 is then applied in the regression between modeledEc
LUC and modeled initial biomass among different DGVMsto calculate constrained Ec
LUC. In this emerging constraintapproach (Fig. 1), the uncertainties in constrained Ec
LUCare a function of the uncertainties in the observed biomassdatasets, the linear regression goodness of fit for the two re-gressions (regressions between Ec
LUC and the initial biomassand between the initial and present biomass) and the slopesof the regressions. The uncertainty in constrained LULCCemissions is calculated as in Stegehuis et al. (2013):
σLULCC =
√α2σ 2
initial_biomass+ σ2res_LULCC , (1)
σinitial_biomass =
√β2σ 2
present_biomass+ σ2res_biomass , (2)
where σLULCC, σinitial_biomass and σpresent_biomass are the un-certainties in constrained Ec
LUC, the uncertainty in initialbiomass and the uncertainty in present biomass; α andσres_LULCC represent the slope and the standard deviation ofthe residuals from the linear regression fit between Ec
LUC andinitial biomass, and β and σres_biomass represent the slope andstandard deviation of the residuals from the linear regressionbetween initial biomass and present biomass.
2.5 Two supplementary methods to constrain EcLUC
using biomass observations
We also tested two supplementary methods to constrainEc
LUC: first, Method S1 using the regression between EcLUC
and present-day biomass from TRENDY models rather thanextrapolating present biomass to biomass in 1901, and thenMethod S2 using 1B (biomass difference between presentbiomass and biomass in 1901 derived from the model simu-lations) instead of a regression between biomass in 1901 andpresent-day biomass to extrapolate the observation-basedbiomass in 1901. In Method S1, the uncertainties in thebiomass observations and in the regression between Ec
LUCand present biomass from the models are used to calculatethe uncertainties in the constrained Ec
LUC. In Method S2, theuncertainties in the biomass observations and the standarddeviation of 1B among the models are used. Ta
ble
3.T
hegl
obal
and
regi
onal
cum
ulat
ive
land
-use
and
land
-cov
erch
ange
(LU
LC
C)
emis
sion
s(P
gC)
duri
ng19
01–2
012
from
orig
inal
TR
EN
DY
mod
els
and
from
the
estim
ates
cons
trai
ned
bydi
ffer
entb
iom
ass
data
sets
with
diff
eren
tmet
hods
tode
fine
defo
rest
atio
ngr
idce
lls.T
hein
terq
uant
ilera
nges
are
show
nin
Tabl
eS1
.
TR
EN
DY
Car
valh
ais
etal
.(20
14)
Liu
etal
.(20
15)
GE
OC
AR
BO
NPa
net
al.(
2011
)(A
vita
bile
etal
.,20
16;
Sant
oro
etal
.,20
15)
med
ian
min
max
Met
hod
AM
etho
dB
Met
hod
CM
etho
dA
Met
hod
BM
etho
dC
Met
hod
AM
etho
dB
Met
hod
CM
etho
dA
Met
hod
BM
etho
dC
Chi
nare
gion
10.7
6.0
19.1
13.8±
4.0
16.0±
4.3
16.3±
4.3
10.5±
3.1
11.1±
3.1
11.1±
3.1
10.0±
4.0
10.5±
4.5
10.5±
4.6
7.3±
2.9
7.6±
2.9
7.7±
2.9
Nor
thA
mer
ica
19.9
8.6
40.8
10.8±
7.1
9.6±
7.0
7.8±
6.7
14.7±
6.9
13.6±
6.8
9.3±
6.6
17.8±
8.3
15.4±
7.7
13.0±
7.6
9.5±
6.4
8.5±
6.4
6.7±
6.4
Sout
han
dC
entr
alA
mer
ica
42.6
33.5
81.4
44.4±
17.8
46.4±
18.1
46.8±
18.2
48.3±
17.0
50.1±
17.0
50.6±
17.0
44.5±
16.6
46.4±
16.7
46.8±
16.8
43.1±
17.0
44.8±
17.1
45.1±
17.2
Wes
tern
Eur
ope
3.8
1.2
5.3
3.6±
0.8
3.2±
0.8
3.0±
0.8
4.1±
0.8
3.4±
0.8
3.2±
0.8
5.0±
1.2
3.8±
0.9
3.4±
0.8
3.6±
0.8
3.2±
0.8
3.0±
0.8
Trop
ical
Afr
ica
21.8
15.8
57.8
24.6±
10.3
28.2±
11.4
28.6±
11.5
31.4±
8.8
36.3±
9.4
36.9±
9.4
22.7±
12.8
26.2±
14.6
26.4±
16.1
23.8±
10.3
27.5±
11.4
27.8±
11.5
The
form
erSo
viet
Uni
on10
.57.
233
.010
.9±
6.7
10.7±
6.7
11.3±
6.8
14.2±
6.5
14.3±
6.5
14.9±
6.5
14.7±
6.6
13.0±
6.5
13.5±
6.5
10.1±
6.2
9.8±
6.2
10.2±
6.2
Sout
han
dSo
uthe
astA
sia
21.8
9.6
46.6
37.2±
14.4
33.6±
13.3
38.5±
14.8
27.8±
8.5
24.1±
7.9
27.9±
8.3
32.9±
10.1
29.0±
9.6
33.6±
10.4
15.1±
9.5
13.3±
8.9
15.5±
9.6
Paci
ficde
velo
ped
regi
on3.
6−
6.0
18.6
6.0±
4.4
5.4±
4.2
6.0±
4.3
7.1±
3.5
6.0±
3.4
5.6±
3.3
18.0±
3.1
16.4±
2.9
14.3±
3.0−
1.7±
2.6−
1.9±
2.6−
2.0±
2.6
Nor
thA
fric
aan
dth
eM
iddl
eE
ast
1.7−
0.6
6.0
1.1±
0.6
0.6±
0.6
0.8±
0.6
4.3±
1.1
3.1±
0.8
3.5±
0.9
4.5±
4.5
3.0±
3.0
3.1±
3.7−
0.1±
0.5−
0.2±
0.5−
0.1±
0.5
Glo
bal
148
9427
315
2±
4915
4±
5015
9±
5116
1±
4016
2±
3916
3±
3916
5±
4616
0±
4516
1±
4711
9±
3712
1±
3812
2±
38
www.biogeosciences.net/14/5053/2017/ Biogeosciences, 14, 5053–5067, 2017
5060 W. Li et al.: Land-use and land-cover change carbon emissions
10 20 30 40
05
101520
y = 0.37x+1.48 r = 0.262
China region
10 30 50 70
0
20
40
y = 0.42x+0.16 r = 0.702
North America
100 150 200 250 30010
30
50
70
90
y = 0.21x+12.09 r = 0.612
South and Central America
2 6 10 14
0
2
4
6
8
y = 0.13x+2.45 r = 0.092
Western Europe
20 60 100 1400
20
40
60
y = 0.37x-0.94 r = 0.762
Tropical Africa
10 30 50 70 900
10
20
30
y = 0.29x+2.57 r = 0.902
The former Soviet Union
20 50 80 110 140
0
20
40
60
y = 0.36x-2.02 r = 0.582
South and Southeast Asia
5 15 25 3510
0
10
20
y = 0.44x-4.72 r = 0.492
Pacific developed region
0 2 4 62
2
6
y = 0.63x-0.30 r2= 0.48
North Africa and Middle East
CLM4.5JSBACHJULES3.2LPJLPJ-GUESSLPX-BernORCHIDEEVISITOCN
300 500 700 9000
50
100
150
200
250
300
y = 0.31x+11.80 r = 0.662
Globe
Cum
ulat
ive
LULC
C e
mis
sion
s (P
g C
)
Biomass in 1901 (Pg C)
Figure 3. Relationship between biomass in 1901 and cumulative land-use and land-cover change (LULCC) emissions during 1901–2012across the nine TRENDY-v2 models. The black solid line is the linear regression line. The vertical green solid line indicates the reconstructedbiomass in 1901 from Carvalhais et al. (2014) by applying Method A (the increase in cropland in HYDE v3.1 data from forest; see Figs. S4and S5 for the results of Method B and Method C) to define deforestation grid cells. The orange solid horizontal line indicates the cumulativeLULCC emissions constrained by reconstructed biomass in 1901. Dashed lines represent 1σ uncertainties. The probability density functionof the constrained cumulative LULCC emissions is shown on the right.
3 Results
3.1 Forest area change and cumulative LULCCemissions in DGVMs
As expected, a general decrease in forest area is found be-tween 1901 and 2012, especially in regions subject to ex-tensive deforestation over the last decades, namely Southand Central America, South and Southeast Asia and tropi-cal Africa (Fig. 2), which is in support of our methods ofdefining deforestation grid cells, although the forest area insome regions differs substantially across DGVMs. Differ-ences in forest area are large in tropical Africa, North Amer-ica and the former Soviet Union, while they are smaller inSouth and Central America and South and Southeast Asia(Fig. 2). There are several reasons for these differences inforest area: (1) the models have different initial distributionsof PFTs (the TRENDY-v2 protocol only prescribed the sameinitial area of natural vegetation, but did not specify the PFTsthat compose natural vegetation); (2) some models consideronly net LULCC, but others have gross LULCC includingsome sub-grid transitions (Table 2; see a comparison usingthe JSBACH model; Wilkenskjeld et al., 2014); (3) and themodels have different treatments for changing pasture areas(either proportional from natural vegetation or preferentialfrom natural grasslands). In North America, the China re-gion and Western Europe, the forest area decreased in the first
half of the 20th century and then increased in recent decades.Yet, the magnitude of the increase is smaller than that of theprevious decrease in these regions, and the global average isnet forest loss between 1901 and 2012 (ranging from 2.3 to16.8 Mkm2 across the nine models).
EcLUC from the nine DGVMs between 1901 and 2012
range from 1.7 PgC (−0.6 to 6.0; median and range are pos-itive, indicating a net cumulative flux to the atmosphere) inNorth Africa and the Middle East to 42.6 PgC (33.5 to 81.4)in South and Central America, resulting in a global total of148 PgC (94 to 273; Table 3). Tropical Africa and South andSoutheast Asia have the second-largest Ec
LUC of 21.8 (15.8 to57.8) and 21.8 PgC (9.6 to 46.6), respectively. Although af-forestation and reforestation occurred in North America afteraround 1960 and in China after 2000 (Fig. 2), Ec
LUC in thesetwo regions have been positive since 1901, with median val-ues of 19.9 and 10.7 PgC, respectively (Table 3).
3.2 Relationship between cumulative LULCCemissions and initial biomass
We found a positive linear relationship between EcLUC and
initial biomass in the deforestation grid cells of each modelon a global scale and in the regions considered (Fig. 3). Thecoefficients of determination (r2) are 0.61, 0.58 and 0.76 inSouth and Central America, South and Southeast Asia andtropical Africa, respectively. Due to stable or slightly in-
Biogeosciences, 14, 5053–5067, 2017 www.biogeosciences.net/14/5053/2017/
W. Li et al.: Land-use and land-cover change carbon emissions 5061
creasing forest area (Fig. 2), the correlation between initialbiomass and Ec
LUC is small in Western Europe (Fig. 3). Theslopes of the relationships between Ec
LUC and initial biomassshown in Fig. 3 range from 0.13 PgC PgC−1 in Western Eu-rope to 0.63 PgC PgC−1 in North Africa and the Middle East.In tropical regions with intensive LULCC, the slope is simi-lar between South and Southeast Asia (0.36 PgC PgC−1) andtropical Africa (0.37 PgC PgC−1), but lower in South andCentral America (0.21 PgC PgC−1). These slopes reflect thesensitivity of cumulative carbon loss to initial biomass car-bon stock. They are mainly influenced by the fraction of de-forested area relative to the initial forest area in each region,which explains 46 % of the variations in the slopes acrossregions (Fig. S3). Differences in biomass density acrossregions and in the use of gross or net transitions amongDGVMs (Table 2) also contribute to variations in slopes.
3.3 Cumulative LULCC emissions constrained bypresent-day biomass observations
There is also a strong positive relationship between initialbiomass in 1901 and present-day biomass in grid cells thathave experienced deforestation (Fig. 4). The r2 of this re-gression is higher than 0.92 in most regions, except in NorthAmerica and the China region (0.89 and 0.76, respectively).The regression between present-day and initial biomass wasapplied to extrapolate current observation-based biomassback to the year 1901. The extrapolated biomass in 1901 ishigher than that in the present day, mainly due to a larger for-est area, although it is difficult to discriminate other effects,such as CO2 fertilization, that might have increased biomassbetween 1901 and 2012.
Using the chain of emerging constraints between present-day and initial biomass (Fig. 4) and between Ec
LUC and ini-tial biomass (Fig. 3), with all uncertainties being propagated(Eqs. 1 and 2), we were able to constrain Ec
LUC during 1901–2012 by biomass observations (Figs. 3, S4, S5, Table 3). TheEc
LUC value constrained by the biomass dataset of Carvalhaiset al. (2014) is 155± 50 PgC (mean and 1σ Gaussian error)and this estimate is robust to the choice of the methods to de-fine deforestation grid cells in biomass datasets (constrainedEc
LUC= 152± 49, 154± 50 and 159± 51 PgC for Method A,Method B and Method C, respectively). The difference be-tween the global constrained Ec
LUC and the median value oforiginal Ec
LUC (148 PgC) from TRENDY DGVMs is not sig-nificant, suggesting that the median model estimate is inde-pendently verified by biomass observations. Still, some mod-els that are inconsistent with the observations can be identi-fied (Fig. 3).
The uncertainties reported in our constrained estimate ofEc
LUC include uncertainties in the biomass observations andin the scatter of the two regressions (Figs. 3, 4) used to con-struct the emerging constraint. The uncertainties in the con-strained Ec
LUC are still relatively large, resulting from thelarge uncertainties in the biomass observations. However, it
should be noted that we summed the biomass uncertaintyin each deforestation grid cell to give the regional biomassuncertainty, which gives a maximum uncertainty with a po-tential assumption that the uncertainties in all grid cells arefully correlated. In reality, the regional biomass uncertaintyshould be lower, thus leading to lower uncertainty in con-strained Ec
LUC. However, it is difficult to estimate the errorcorrelations of observation-based biomass between differentgrid cells at this stage.
Although the constrained global EcLUC value is only 7 PgC
higher than the median of the original DGVM ensemble (Ta-ble 3), larger differences can be found on a regional scale(Fig. 5). Constrained Ec
LUC estimates are higher than the orig-inal modeled values in South and Southeast Asia, tropicalAfrica and South and Central America (Table 3). For exam-ple, the constrained Ec
LUC value is 37.2± 14.4 PgC in Southand Southeast Asia compared to the original TRENDY me-dian value of 21.8 PgC (range of 9.6 to 46.6 PgC) for that re-gion. The constrained emissions are also higher in the Chinaregion and the Pacific developed region compared to theprior median value (see Table 3). A significantly large re-duction in Ec
LUC through the emerging constraint is found inNorth America because of the lower biomass amount fromobservation-based datasets than from DGVMs. The originalmedian Ec
LUC value of that region is 19.9 PgC (range of 8.6to 40.8 PgC), while the constrained result is 10.8± 7.1 PgC.Constrained Ec
LUC are also lower than original estimates inWestern Europe, North Africa and the Middle East, althoughtheir contributions to the global total emissions are very small(Table 3).
Alternative estimates of EcLUC constrained by three other
biomass datasets (Liu et al., 2015; GEOCARBON, Avitabileet al., 2016; Santoro et al., 2015; Pan et al., 2011) are pro-vided in Fig. 6 and Table 3. In general, the constrained Ec
LUCusing biomass maps from Liu et al. (2015) and GEOCAR-BON (Avitabile et al., 2016; Santoro et al., 2015) are ratherconsistent (on average only 4.5 % higher) with those fromCarvalhais et al. (2014), implying the robustness of our es-timates. The biomass dataset from Pan et al. (2011) leads tolower LULCC emission estimates on a global scale, mainlydue to a lower estimate in South and Southeast Asia (Ta-ble 3) compared to the other products. In the Pacific devel-oped region, GEOCARBON-based estimates (Avitabile etal., 2016; Santoro et al., 2015) are much higher than thosefrom Carvalhais et al. (2014) because the latter has a gap inthe biomass map in the southern part of Australia (Carval-hais et al., 2014). In Fig. 6, we show the original Ec
LUC fromTRENDY DGVMs as quantiles because we do not knowwhether they follow a normal distribution; to be comparable,the interquantiles of the constrained Ec
LUC are also shown.The interquantile range of constrained Ec
LUC is larger thanthat of the original Ec
LUC (Fig. 6). This, however, does notmean that our emerging constraint method is not effective,but that the relatively large uncertainty in the constrainedEc
LUC is propagated from the biomass observation uncer-
www.biogeosciences.net/14/5053/2017/ Biogeosciences, 14, 5053–5067, 2017
5062 W. Li et al.: Land-use and land-cover change carbon emissions
10 20 30 4010
20
30
40
y = 0.86x+10.10 r = 0.762
China region
25 50 75
20
40
60
80
y = 1.07x+3.25 r = 0.892
North America
100 200 300
100
200
300
y = 0.92x+33.85 r = 0.922
South and Central America
5 10 15
5
10
15
y = 1.03x+1.27 r = 0.942
Western Europe
20 60 100 14020
60
100
140
y = 1.23x-1.83 r = 0.972
Tropical Africa
0 50 1000
50
100
y = 0.95x+3.84 r = 0.952
The former Soviet Union
50 100
50
100
y = 1.12x+2.12 r = 0.962
South and Southeast Asia
20 4010
20
30
40
y = 1.07x+0.18 r = 0.942
Pacific developed region
0.0 2.5 5.00
2
4
6
y = 1.49x-0.21 r = 0.962
North Africa and Middle East
CLM4.5JSBACHJULES3.2LPJLPJ-GUESSLPX-BernORCHIDEEVISITOCN
Bio
mas
s in
190
1 (P
g C
)
Present biomass (Pg C)
Figure 4. The relationship between initial biomass in 1901 and present biomass (average of biomass from 2000 to 2012) across the TRENDY-v2 models for each region. Note that both biomass in 1901 and present biomass are from TRENDY models, not the observations. Dashedline is the 1 : 1 line.
tainty, which is about one-third of the mean biomass at theglobal level (Carvalhais et al., 2014).
The global constrained EcLUC value obtained by using the
two supplementary methods is almost identical to that fromour original method in Fig. 1 (see an example in Fig. S6).The difference in Ec
LUC between the supplementary and orig-inal methods at the global level is < 1 % for all biomass ob-servation datasets (Carvalhais et al., 2014; Liu et al., 2015;GEOCARBON, Avitabile et al., 2016; Santoro et al., 2015;Pan et al., 2011) and all methods to select LULCC grid cells(Method A, B and C). This suggests that our constrained re-sults are very robust. The change in the uncertainty in globalconstrained Ec
LUC is also very small (< 2 %) because mostof the uncertainties are from the biomass observations (seeDiscussion) and the regression between Ec
LUC and biomass(see r2 in Fig. 3), rather than from converting present-daybiomass to biomass in 1901 (see r2 in Fig. 4). The differ-ence in regional Ec
LUC between different constraint methodsis relatively larger (12 % on average), but the difference re-mains very small in tropical regions (∼ 1 %). However, wenote that the results from the two supplementary methods(Method S1 and S2) should be cautiously treated. First, be-cause Ec
LUC are related to the biomass that has been affectedsince the start of the land-use perturbation, only biomassin 1901 (rather than that left out of land use in the 2000s)
in LULCC-affected grid cells is logically related to histori-cal Ec
LUC. Thus, converting present-day biomass to biomassin 1901 (the original method; Fig. 1) is a more direct andprocess-justified approach compared to regressing present-day biomass versus Ec
LUC (Method S1), which is not justi-fied by a logical mechanism. Second, using 1B in MethodS2 is not a perfect solution to extrapolate biomass in 1901from present-day biomass because the change in biomass isnot solely impacted by land-use change. The interactions be-tween biomass and climate conditions, disturbances and nu-trient limitation are also very important in DGVMs. For ex-ample, historical LULCC may reduce biomass over LULCC-affected regions by replacing forests with croplands. On thecontrary, the CO2 fertilization effects may increase biomassover LULCC and non-LULCC regions. Therefore, 1B re-flects a mixed effect of different factors, not a sole responseto LULCC. In addition, as 1B has a higher relative uncer-tainty among models (∼ 53 % at the global level), using theregression (r2 > 0.92 in seven regions; Fig. 4) to calculatebiomass in 1901 could include relatively less noisy informa-tion than using 1B.
Biogeosciences, 14, 5053–5067, 2017 www.biogeosciences.net/14/5053/2017/
W. Li et al.: Land-use and land-cover change carbon emissions 5063
0 20 400
20
40
(a)
0 20 400
20
40
(b)
0 20 400
20
40
(c)China regionNorth AmericaSouth and Central AmericaWestern EuropeTropical AfricaThe former Soviet UnionSouth and Southeast AsiaPacific developed regionNorth Africa and Middle East
Con
stra
ined
cum
ulat
ive
LULC
C e
mis
sion
s (P
g C
)
Original cumulative LULCC emissions (Pg C)
Figure 5. Comparisons between the original TRENDY land-use andland-cover change (LULCC) emissions and the cumulative LULCCemissions constrained by the biomass dataset from Carvalhais etal. (2014). Panels (a), (b) and (c) are the results from Method A,Method B and Method C, respectively. The original TRENDY emis-sions are shown as the median value of all models. Dashed line isthe 1 : 1 line.
4 Discussion
Our approach to constraining EcLUC from an ensemble of
DGVMs provides a best estimate that is between those fromtwo bookkeeping models (∼ 130 PgC from Houghton et al.,2012, and 212 PgC for the default dataset from Hansis etal., 2015). Although the bookkeeping model from Hansiset al. (2015) was driven by the same agricultural land-usemaps as the TRENDY models (the model of Houghton etal., 2012, uses FRA/FAO data), the Ec
LUC value from Han-sis et al. (2015) is different from that constrained fromthe DGVMs. Differences in estimates between DGVMs andbookkeeping models have been attributed to different defini-tions of LULCC emissions (Pongratz et al., 2014; Stockerand Joos, 2015). Indeed, LULCC emissions from DGVMsimulations in TRENDY include the “missed sink capacityin the deforested area” (Gasser and Ciais, 2013; Pongratzet al., 2014), and so, all else being equal, should simulatehigher emissions than bookkeeping models, which do notinclude this term. However, bookkeeping models take for-est degradation into account, while this process is ignoredin DGVMs. Bookkeeping models also represent shifting cul-tivation (resulting in larger sub-grid-scale gross land transi-tions as opposed to net transitions) and wood harvest; theseare processes that are accounted for in only a subset of theTRENDY models (see Table 2). In addition to different driv-ing LULCC area data, differences between the two book-keeping models were discussed by Hansis et al. (2015); for
OriginalTRENDY
Carvalhais et al. Liu et al. GEOCARBON Pan et al.0
50
100
150
200
250
Cum
ulat
ive
LULC
C e
mis
sion
s (P
g C
)
Met
hod
A
Met
hod
B
Met
hod
C
All
met
hods
Met
hod
A
Met
hod
B
Met
hod
C
All
met
hods
Met
hod
A
Met
hod
B
Met
hod
C
All
met
hods
Met
hod
A
Met
hod
B
Met
hod
C
All
met
hods
Figure 6. The global cumulative land-use and land-cover change(LULCC) emissions during 1901–2012 from original TRENDYmodels and from the estimates constrained by different biomassdatasets with different methods to define deforestation grid cells.“All methods” represents the ensemble mean and uncertainty in theconstrained results from Method A, Method B and Method C foreach biomass dataset. The whisker–box plot represents the mini-mum and maximum values, 25th and 75th percentiles and the me-dian value of original TRENDY models. In the bar plot for the con-strained estimates, the red line represents the 1σ Gaussian errors;the black ticks represent the 25th and 75th percentiles.
example, Houghton et al. (2012) assumed a preferential al-location of pastures on natural grasslands, while Hansis etal. (2015) assumed a proportional allocation of both croplandand pasture on all available natural vegetation types.
We are aware that our truncated diagnostic of a set of de-forestation grid cells, instead of grid cells affected by allLULCC types, is an underestimate of the total area subjectto LULCC because we ignore grid cells that experiencedland-use transitions between non-forest vegetation only (e.g.,only conversions from grasslands to cropland happening ina grid cell). However, the conversion of forest to croplandsand pasture dominates the total net LULCC flux (Houghton,2003, 2010), while the contribution of transitions betweennon-forest vegetation and agriculture to Ec
LUC is compara-tively small (Fig. S1). In fact, the annual LULCC emissionfrom deforestation was estimated to be 2.2 PgC yr−1 duringthe 1990s, and the total emissions from other activities (e.g.,afforestation, reforestation, non-forest transitions) are nearlyneutral (Houghton, 2003).
The lack of direct biomass observations at the initial stateforces us to hindcast biomass in 1901 based on present-dayobservations; this is an extrapolation that also comes withuncertainties. Some of the observed biomass datasets onlycover forests, and satellite measurements usually quantifyaboveground biomass carbon stocks and not total biomassstocks (Table 1). In addition, the regression of modeledbiomass between 1901 and 2000–2012 (average) to extrap-olate the biomass amount in 1901 is only a statistical ap-proach. This regression cannot be mechanistically explainedbecause its slope and intercept are impacted by multiple fac-tors in the models like land clearing, secondary vegetationregrowth, CO2 fertilization, climate, disturbances and the nu-
www.biogeosciences.net/14/5053/2017/ Biogeosciences, 14, 5053–5067, 2017
5064 W. Li et al.: Land-use and land-cover change carbon emissions
trient limitation on biomass. Despite these uncertainties, thehigh coefficient of determination in the regression increasesour confidence in the biomass extrapolation to 1901. For agiven biomass dataset, the choice of a method for definingdeforestation grid cells (Method A, Method B and MethodC) has a very small influence on our results (Table 3).
LULCC carbon emissions are influenced not only bychanges in biomass, but also by how these are prescribed inthe model to influence posterior changes in detrital and soilorganic carbon pools. However, LULCC emissions are domi-nated by changes in biomass. For example, LULCC results ina net carbon loss of 110 PgC in biomass during 1850–1990,accounting for 89 % of the total Ec
LUC (Houghton, 1999). Thesoil carbon changes after LULCC is also indirectly impactedby initial biomass, since the dead roots and remaining above-ground debris turn into soil organic carbon after land clear-ing, which takes longer to return into the atmosphere. In addi-tion, it is not necessary to account for all factors when apply-ing an emergent constraint approach (e.g., Cox et al., 2013;Kwiatkowski et al., 2017; Wenzel et al., 2016). The regres-sion between Ec
LUC and biomass in 1901 in the models in ourstudy is satisfying (e.g., r2
= 0.66 on a global scale; Fig. 3)to constrain Ec
LUC through biomass observations.The required model outputs for carbon stocks and fluxes
in the TRENDY project are not PFT specific; only the meanPFT-mixed variables in each grid cell are required. Such anaggregation prevents a rigorous separation of biomass be-tween forest and other biomes in each grid cell. It was thusimpossible for us to calculate individual contributions of dif-ferent LULCC types to the overall LULCC emissions, whichinduces uncertainties when matching model results with ob-served forest biomass distributions (e.g., only forest biomassin datasets from GEOCARBON; Avitabile et al., 2016; San-toro et al., 2015; Pan et al., 2011). Therefore, we suggestthat the next generation of DGVM comparisons report PFT-specific carbon stock and fluxes, and other model intercom-parison exercises should follow suit. The approach of us-ing multiple biomass observation datasets to constrain theLULCC emissions could also be applied in other model-ing projects, such as Coupled Model Intercomparison ProjectPhase 5 (CMIP5) and CMIP6.
Currently, the uncertainties in the satellite-based biomassdatasets are relatively large (e.g., 38 % on average in the trop-ics at the pixel level (< 1 km); Saatchi et al., 2011). This in-troduces uncertainties in the constrained cumulative LULCCemissions, depending on the forest types and biomass range.For example, on average on the global scale, the uncertaintyin the resolution of DGVM grid cells (0.5◦× 0.5◦) is aboutone-third of the mean biomass (Carvalhais et al., 2014) andthe relative uncertainty is smaller for high biomass areas inthe tropics (Avitabile et al., 2016; Saatchi et al., 2011).
The main sources of uncertainties in satellite-basedbiomass datasets depend on the specific product, the spa-tial resolution of the datasets and the methodology used tovalidate the data. For instance, in the case of radar remote
sensing used for biomass mapping in Northern Hemisphereboreal and temperate forests, the uncertainty is largely due tothe sensitivity of the signal to properties other than vegetationstructure (e.g., moisture), the influence of non-forest vegeta-tion on the signal (especially in fragmented landscapes; San-toro et al., 2015) and uncertainties in the additional datasets(allometric databases, land cover) used for the conversion ofsatellite measurements to biomass estimates (Thurner et al.,2014). At the pixel level and modeling grid cells, uncertain-ties may also be strongly influenced by the quality and sizeof the inventory data used for validation and the significantmismatch between pixel area and the plot data, as well as thedifference between the dates of satellite and ground observa-tions (Saatchi et al., 2015, 2011; Thurner et al., 2014).
Moreover, the satellite-derived biomass datasets used inthis study represent different dates. The tropical biomassproducts represent the circa 2000 status of forests, whereasthe boreal and temperate biomass maps are based on space-borne radar data from the year 2010. These differences inthe date of observations introduce additional uncertainty inthe biomass estimates due to changes in forest cover fromthe disturbance, recovery and land-use activities (Hurtt et al.,2011) occurring annually and regionally.
However, in boreal, temperate and in tropical regions, theestimated relative uncertainties were lowest in high biomassareas (Avitabile et al., 2016; Thurner et al., 2014), whichdominate the contribution to our results. Moreover, the rel-atively high accuracy of biomass datasets when aggregatedto modeling grid cells from higher-resolution maps (< 1 km;Saatchiet al., 2011; Thurner et al., 2014) suggests that thebiomass datasets implemented in our study provide a realis-tic representation of carbon stocks to constrain the historicalcumulative LULCC emissions from vegetation.
5 Conclusions
Uncertainties in LULCC carbon emissions are relativelylarge compared to other terms in the global carbon bud-get. The wide spread is partly due to the differences inmodel structure but also because of the difficulty in con-straining models by observations of LULCC, particularlyemissions resulting from deforestation. We propose an ob-servationally constrained global cumulative LULCC emis-sion of 155± 50 PgC during 1901 and 2012. Although theconstrained cumulative LULCC emissions are close to theunconstrained ones from models, our study offers an eval-uation of the modeling results using the observation-basedbiomass. More importantly, we combine the uncertainties inthe regressions from state-of-the-art models with uncertain-ties in multiple observation-based biomass datasets and givea constrained Ec
LUC with a 1σ Gaussian uncertainty. The ideaof an emergent constraint approach is to give a more accu-rate estimate and/or reduced uncertainty in an unknown vari-able by combining a heuristic relationship between two mod-
Biogeosciences, 14, 5053–5067, 2017 www.biogeosciences.net/14/5053/2017/
W. Li et al.: Land-use and land-cover change carbon emissions 5065
eled variables (an observable and an unknown one) with ac-tual observations of the observable variable. Thus, our studyshows (1) that there is a heuristic relationship between initialbiomass and Ec
LUC among different models, (2) that availablebiomass observation data independently confirm the medianof modeled emission estimates and (3) that more accuratebiomass data in the future would allow some of the mod-eled estimates of emissions to be falsified. Although the un-certainties in current observation-based biomass datasets arerelatively high, as more accessible and accurate observationdata become available, many data-driven opportunities arebeing created to improve the accuracy of DGVM predictions.
Data availability. Different biomass datasets used in this studycan be downloaded based on information in their original publica-tions. Specifically, the biomass dataset of Carvalhais et al. (2014)can be downloaded from MPI BGI Data Portal: https://www.bgc-jena.mpg.de/geodb/projects/Home.php; The biomass datasetof Liu et al. (2015) can be downloaded from http://www.wenfo.org/wald/global-biomass/; The biomass dataset of GEO-CARBON (Avitabile et al., 2016; Santoro et al., 2015) can bedownloaded from http://www.wur.nl/en/Expertise-Services/Chair-groups/Environmental-Sciences/; The regional biomass of Pan etal. (2011) can be found in Table 2 in their paper. The outputs(biomass-constrained cumulative LULCC emissions) of this studyare provided in Table 3.
The Supplement related to this article is available onlineat https://doi.org/10.5194/bg-14-5053-2017-supplement.
Competing interests. The authors declare that they have no conflictof interest.
Acknowledgements. Wei Li, Chao Yue, Thomas A. M. Pugh andAlmut Arneth were supported by the project LUC4C funded bythe European Commission (grant no. 603542). Philippe Ciaisand Shushi Peng acknowledge support from the European Re-search Council through Synergy grant ERC-2013-SyG-610028“IMBALANCE-P”. Julia Pongratz, Julia E. M. S. Nabel andRasoul Yousefpour were supported by the German ResearchFoundation’s Emmy Noether Program (PO 1751/1-1). Ben-jamin D. Stocker was supported by the Swiss National ScienceFoundation and FP7 funding through project EMBRACE (282672).Anna B. Harper was supported by the UK Natural EnvironmentResearch Council Joint Weather and Climate Research Programme.Martin Thurner acknowledges funding from the Vetenskapsrådet(grant no. 621-2014-4266 of the Swedish Research Council).The biomass maps and model outputs can be freely accessed byfollowing the instructions in the original publications. All thebiomass-constrained LULCC emission data can be freely obtainedfrom Wei Li (email: wei.li@lsce.ipsl.fr).
Edited by: Christopher A. WilliamsReviewed by: two anonymous referees
References
Avitabile, V., Herold, M., Heuvelink, G. B. M., Lewis, S. L.,Phillips, O. L., Asner, G. P., Armston, J., Ashton, P. S., Banin,L. F., Bayol, N., Berry, N. J., Boeckx, P., de Jong, B. H. J., De-Vries, B., Girardin, C. A. J., Kearsley, E., Lindsell, J. A., Lopez-Gonzalez, G., Lucas, R., Malhi, Y., Morel, A., Mitchard, E. T.A., Nagy, L., Qie, L., Quinones, M. J., Ryan, C. M., Ferry, S.J. W., Sunderland, T., Laurin, G. V., Gatti, R. C., Valentini, R.,Verbeeck, H., Wijaya, A., Willcock, S., Asthon, P., Banin, L. F.,Bayol, N., Berry, N. J., Boeckx, P., de Jong, B. H. J., DeVries, B.,Girardin, C. A. J., Kearsley, E., Lindsell, J. A., Lopez-Gonzalez,G., Lucas, R., Malhi, Y., Morel, A., Mitchard, E. T. A., Nagy,L., Qie, L., Quinones, M. J., Ryan, C. M., Slik, F., Sunderland,T., Vaglio Laurin, G., Valentini, R., Verbeeck, H., Wijaya, A.,and Willcock, S.: An integrated pan-tropical biomass map usingmultiple reference datasets, Glob. Change Biol., 22, 1406–1420,https://doi.org/10.1111/gcb.13139, 2016.
Baccini, A., Goetz, S. J., Walker, W. S., Laporte, N. T., Sun, M.,Sulla-Menashe, D., Hackler, J., Beck, P. S. A., Dubayah, R.,Friedl, M. A., Samanta, S., and Houghton, R. A.: Estimatedcarbon dioxide emissions from tropical deforestation improvedby carbon-density maps, Nature Climate Change, 2, 182–185,https://doi.org/10.1038/nclimate1354, 2012.
Best, M. J., Pryor, M., Clark, D. B., Rooney, G. G., Essery, R. . L.H., Ménard, C. B., Edwards, J. M., Hendry, M. A., Porson, A.,Gedney, N., Mercado, L. M., Sitch, S., Blyth, E., Boucher, O.,Cox, P. M., Grimmond, C. S. B., and Harding, R. J.: The JointUK Land Environment Simulator (JULES), model description –Part 1: Energy and water fluxes, Geosci. Model Dev., 4, 677–699,https://doi.org/10.5194/gmd-4-677-2011, 2011.
Boden, T. A., Marland, G., and Andres, R. J.: Global, regional, andnational fossil-fuel CO2 emissions, Carbon Dioxide Inf. Anal.Center, Oak Ridge Natl. Lab. USA, Oak Ridge, TN Dep. Energy,https://doi.org/10.3334/CDIAC/00001, 2013.
Carvalhais, N., Forkel, M., Khomik, M., Bellarby, J., Jung, M.,Migliavacca, M., Mu, M., Saatchi, S., Santoro, M., Thurner, M.,Weber, U., Ahrens, B., Beer, C., Cescatti, A., Randerson, J. T.,and Reichstein, M.: Global covariation of carbon turnover timeswith climate in terrestrial ecosystems, Nature, 514, 213–217,https://doi.org/10.1038/nature13731, 2014.
Clark, D. B., Mercado, L. M., Sitch, S., Jones, C. D., Gedney, N.,Best, M. J., Pryor, M., Rooney, G. G., Essery, R. L. H., Blyth,E., Boucher, O., Harding, R. J., Huntingford, C., and Cox, P.M.: The Joint UK Land Environment Simulator (JULES), modeldescription – Part 2: Carbon fluxes and vegetation dynamics,Geosci. Model Dev., 4, 701–722, https://doi.org/10.5194/gmd-4-701-2011, 2011.
Cox, P. M., Pearson, D., Booth, B. B., Friedlingstein, P., Hunting-ford, C., Jones, C. D., and Luke, C. M.: Sensitivity of tropicalcarbon to climate change constrained by carbon dioxide variabil-ity, Nature, 494, 341–344, 2013.
Gasser, T. and Ciais, P.: A theoretical framework for the net land-to-atmosphere CO2 flux and its implications in the definition of
www.biogeosciences.net/14/5053/2017/ Biogeosciences, 14, 5053–5067, 2017
5066 W. Li et al.: Land-use and land-cover change carbon emissions
“emissions from land-use change”, Earth Syst. Dynam., 4, 171–186, https://doi.org/10.5194/esd-4-171-2013, 2013.
Hansis, E., Davis, S. J., and Pongratz, J.: Relevance ofmethodological choices for accounting of land use changecarbon fluxes, Global Biogeochem. Cy., 29, 1230–1246,https://doi.org/10.1002/2014GB004997, 2015.
Harris, N. L., Brown, S., Hagen, S. C., Saatchi, S. S.,Petrova, S., Salas, W., Hansen, M. C., Potapov, P. V.,and Lotsch, A.: Baseline map of carbon emissions fromdeforestation in tropical regions, Science, 336, 1573–1576,https://doi.org/10.1126/science.1217962, 2012.
Houghton, R. A.: The annual net flux of carbon to the atmospherefrom changes in land use 1850–1990, Tellus B, 51, 298–313,https://doi.org/10.1034/j.1600-0889.1999.00013.x, 1999.
Houghton, R. A.: Revised estimates of the annual net fluxof carbon to the atmosphere from changes in land useand land management 1850–2000, Tellus B, 55, 378–390,https://doi.org/10.1034/j.1600-0889.2003.01450.x, 2003.
Houghton, R. A.: How well do we know the flux ofCO2 from land-use change?, Tellus B, 62, 337–351,https://doi.org/10.1111/j.1600-0889.2010.00473.x, 2010.
Houghton, R. A., House, J. I., Pongratz, J., Van Der Werf, G. R.,Defries, R. S., Hansen, M. C., Le Quéré, C., and Ramankutty,N.: Carbon emissions from land use and land-cover change, Bio-geosciences, 9, 5125–5142, https://doi.org/10.5194/bg-9-5125-2012, 2012.
Hurtt, G. C., Chini, L. P., Frolking, S., Betts, R. A., Feddema, J.,Fischer, G., Fisk, J. P., Hibbard, K., Houghton, R. A., Janetos,A., Jones, C. D., Kindermann, G., Kinoshita, T., Klein Gold-ewijk, K., Riahi, K., Shevliakova, E., Smith, S., Stehfest, E.,Thomson, A., Thornton, P., van Vuuren, D. P., and Wang, Y. P.:Harmonization of land-use scenarios for the period 1500–2100:600 years of global gridded annual land-use transitions, woodharvest, and resulting secondary lands, Climate Change, 109,117–161, https://doi.org/10.1007/s10584-011-0153-2, 2011.
Ito, A. and Inatomi, M.: Use of a process-based model for as-sessing the methane budgets of global terrestrial ecosystemsand evaluation of uncertainty, Biogeosciences, 9, 759–773,https://doi.org/10.5194/bg-9-759-2012, 2012.
Kato, E., Kinoshita, T., Ito, A., Kawamiya, M., and Yama-gata, Y.: Evaluation of spatially explicit emission scenarioof land-use change and biomass burning using a process-based biogeochemical model, J. Land Use Sci., 8, 104–122,https://doi.org/10.1080/1747423X.2011.628705, 2013.
Klein Goldewijk, K., Beusen, A., Van Drecht, G., and De Vos,M.: The HYDE 3.1 spatially explicit database of human-inducedglobal land-use change over the past 12,000 years, GlobalEcol. Biogeogr., 20, 73–86, https://doi.org/10.1111/j.1466-8238.2010.00587.x, 2011.
Krinner, G., Viovy, N., de Noblet-Ducoudré, N., Ogée, J., Polcher,J., Friedlingstein, P., Ciais, P., Sitch, S., and Prentice, C. I.:A dynamic global vegetation model for studies of the coupledatmosphere-biosphere system, Global Biogeochem. Cy., 19, 1–33, https://doi.org/10.1029/2003GB002199, 2005.
Kwiatkowski, L., Bopp, L., Aumont, O., Ciais, P., Cox, P.M., Laufkötter, C., Li, Y., and Séférian, R.: Emergent con-straints on projections of declining primary production inthe tropical oceans, Nature Climate Change, 7, 355–358,https://doi.org/10.1038/nclimate3265, 2017.
Le Quéré, C., Moriarty, R., Andrew, R. M., Peters, G. P., Ciais, P.,Friedlingstein, P., and Jones, S. D.: Global carbon budget 2014,Earth Syst. Sci. Data, 7, 47–85, https://doi.org/10.5194/essd-7-47-2015, 2015.
Liu, Y. Y., van Dijk, A. I. J. M., de Jeu, R. A. M., Canadell, J.G., McCabe, M. F., Evans, J. P., and Wang, G.: Recent rever-sal in loss of global terrestrial biomass, Nature Climate Change,5, 470–474, https://doi.org/10.1038/nclimate2581, 2015.
Oleson, K. W., Lawrence, D. M., Bonan, G. B., Drewniak, B.,Huang, M., Koven, C. D., Levis, S., Li, F., Riley, J., Subin, Z.M., Swenson, S. C., Thornton, P. E., Bozbiyik, A., Fisher, R. A.,Heald, C. L., Kluzek, E., Lamarque, J.-F., Lawrence, P. J., Le-ung, L. R., Lipscomb, W., Muszala, S., Ricciuto, D. M., Sacks,W. J., Sun, Y., Tang, J., and Yang, Z.-L.: Technical Descriptionof version 4.5 of the Community Land Model (CLM), 2013.
Pan, Y., Birdsey, R. A., Fang, J., Houghton, R., Kauppi, P. E., Kurz,W. A., Phillips, O. L., Shvidenko, A., Lewis, S. L., Canadell,J. G., Ciais, P., Jackson, R. B., Pacala, S. W., McGuire, A. D.,Piao, S., Rautiainen, A., Sitch, S., and Hayes, D.: A large andpersistent carbon sink in the world’s forests, Science, 333, 988–993, https://doi.org/10.1126/science.1201609, 2011.
Peng, S., Ciais, P., Maignan, F., Li, W., Chang, J., Wang, T., andYue, C.: Sensitivity of land use change emission estimates to his-torical land use and land cover mapping, Global Biogeochem.Cy., 31, 626–643, https://doi.org/10.1002/2015GB005360, 2017.
Pitman, A. J., de Noblet-Ducoudré, N., Cruz, F. T., Davin, E. L.,Bonan, G. B., Brovkin, V., Claussen, M., Delire, C., Ganzeveld,L., Gayler, V., van den Hurk, B. J. J. M., Lawrence, P. J., vander Molen, M. K., Müller, C., Reick, C. H., Seneviratne, S. I.,Strengers, B. J., and Voldoire, A.: Uncertainties in climate re-sponses to past land cover change: First results from the LU-CID intercomparison study, Geophys. Res. Lett., 36, L14814,https://doi.org/10.1029/2009GL039076, 2009.
Pongratz, J., Reick, C. H., Houghton, R. A., and House, J. I.: Ter-minology as a key uncertainty in net land use and land coverchange carbon flux estimates, Earth Syst. Dyn., 5, 177–195,https://doi.org/10.5194/esd-5-177-2014, 2014.
Reick, C. H., Raddatz, T., Brovkin, V., and Gayler, V.: Represen-tation of natural and anthropogenic land cover change in MPI-ESM, Journal of Advances in Modeling Earth Systems, 5, 459–482, https://doi.org/10.1002/jame.20022, 2013.
Saatchi, S., Mascaro, J., Xu, L., Keller, M., Yang, Y., Duffy, P.,Espírito-Santo, F., Baccini, A., Chambers, J., and Schimel, D.:Seeing the forest beyond the trees, Global Ecol. Biogeogr., 24,606–610, https://doi.org/10.1111/geb.12256, 2015.
Saatchi, S. S., Harris, N. L., Brown, S., Lefsky, M., Mitchard, E.T. A., Salas, W., Zutta, B. R., Buermann, W., Lewis, S. L.,Hagen, S., Petrova, S., White, L., Silman, M., and Morel, A.:Benchmark map of forest carbon stocks in tropical regions acrossthree continents, Proc. Natl. Acad. Sci. USA, 108, 9899–904,https://doi.org/10.1073/pnas.1019576108, 2011.
Santoro, M., Beaudoin, A., Beer, C., Cartus, O., Fransson, J. E. S.,Hall, R. J., Pathe, C., Schmullius, C., Schepaschenko, D., Shvi-denko, A., Thurner, M., and Wegmüller, U.: Forest growing stockvolume of the northern hemisphere: Spatially explicit estimatesfor 2010 derived from Envisat ASAR, Remote Sens. Environ.,168, 316–334, https://doi.org/10.1016/j.rse.2015.07.005, 2015.
Sitch, S., Smith, B., Prentice, I. C., Arneth, A., Bondeau, A.,Cramer, W., Kaplan, J. O., Levis, S., Lucht, W., Sykes, M. T.,
Biogeosciences, 14, 5053–5067, 2017 www.biogeosciences.net/14/5053/2017/
W. Li et al.: Land-use and land-cover change carbon emissions 5067
Thonicke, K., and Venevsky, S.: Evaluation of ecosystem dynam-ics, plant geography and terrestrial carbon cycling in the LPJ dy-namic global vegetation model, Glob. Change Biol., 9, 161–185,https://doi.org/10.1046/j.1365-2486.2003.00569.x, 2003.
Sitch, S., Friedlingstein, P., Gruber, N., Jones, S. D., Murray-Tortarolo, G., Ahlström, A., Doney, S. C., Graven, H., Heinze, C.,Huntingford, C., Levis, S., Levy, P. E., Lomas, M., Poulter,B., Viovy, N., Zaehle, S., Zeng, N., Arneth, A., Bonan, G.,Bopp, L., Canadell, J. G., Chevallier, F., Ciais, P., Ellis, R.,Gloor, M., Peylin, P., Piao, S. L., Le Quéré, C., Smith, B.,Zhu, Z., and Myneni, R.: Recent trends and drivers of regionalsources and sinks of carbon dioxide, Biogeosciences, 12, 653–679, https://doi.org/10.5194/bg-12-653-2015, 2015.
Smith, B., Prentice, I. C., and Sykes, M. T.: Representation ofvegetation dynamics in the modelling of terrestrial ecosystems:comparing two contrasting approaches within European climatespace, Global Ecol. Biogeogr., 10, 621–637, 2001.
Stegehuis, A. I., Teuling, A. J., Ciais, P., Vautard, R., and Jung,M.: Future European temperature change uncertainties reducedby using land heat flux observations, Geophys. Res. Lett., 40,2242–2245, https://doi.org/10.1002/grl.50404, 2013.
Stocker, B. D. and Joos, F.: Quantifying differences in land useemission estimates implied by definition discrepancies, EarthSyst. Dynam., 6, 731–744, https://doi.org/10.5194/esd-6-731-2015, 2015.
Stocker, B. D., Feissli, F., Strassmann, K. M., Spahni, R., andJoos, F.: Past and future carbon fluxes from land use change,shifting cultivation and wood harvest, Tellus B, 66, 1–15,https://doi.org/10.3402/tellusb.v66.23188, 2014.
Thurner, M., Beer, C., Santoro, M., Carvalhais, N., Wutzler, T.,Schepaschenko, D., Shvidenko, A., Kompter, E., Ahrens, B.,Levick, S. R., and Schmullius, C.: Carbon stock and density ofnorthern boreal and temperate forests, Global Ecol. Biogeogr.,23, 297–310, https://doi.org/10.1111/geb.12125, 2014.
Wenzel, S., Cox, P. M., Eyring, V., and Friedlingstein, P.: Pro-jected land photosynthesis constrained by changes in theseasonal cycle of atmospheric CO2, Nature, 538, 499–501,https://doi.org/10.1038/nature19772, 2016.
Wilkenskjeld, S., Kloster, S., Pongratz, J., Raddatz, T., and Re-ick, C. H.: Comparing the influence of net and gross an-thropogenic land-use and land-cover changes on the car-bon cycle in the MPI-ESM, Biogeosciences, 11, 4817–4828,https://doi.org/10.5194/bg-11-4817-2014, 2014.
Zaehle, S. and Friend, A. D.: Carbon and nitrogen cycledynamics in the O-CN land surface model: 1. Modeldescription, site-scale evaluation, and sensitivity to pa-rameter estimates, Global Biogeochem. Cy., 24, 1–13,https://doi.org/10.1029/2009GB003521, 2010.
www.biogeosciences.net/14/5053/2017/ Biogeosciences, 14, 5053–5067, 2017
top related